Bump Gradle from 6.8.3 to 8.10.2 and increase the minimum required
Java version from 8 to 11 (SMACK-953).
The switch from Java 8 to 11 caused some Bytecode portability issues
regarding NIO Buffers. Java changed with version 9 the return type of
some subclasses of Buffer to return the specific Buffer type instead
of the Buffer superclass [JDK-4774077]. For example, ByteBuffer.filp()
previously returned Buffer, while it does return ByteBuffer now.
This sensible change was not reflected by the Android API [1], which
means that AnimalSniffer rightfully started to complain that there is
no method "ByteBuffer ByteBuffer.flip()" in Android, there is only
"Buffer ByteBuffer.flip()", and those are incompatible methods on
Java's Bytecode layer.
As workaround, this changes
return charBuffer.flip().toString();
to
((java.nio.Buffer) charBuffer).flip();
return charBuffer.toString();
to restore the Bytecode portability between Android and Java.
Errorprone also got new checks, of which JavaUtilDate and JdkObsolete
are wroth mentioning.
JavaUtilData basically strongly recommends to use Java's newer time
API over java.util.Date. But since Smack was Java 8 until now,
j.u.Date is widely used.
Similar JdkObsolete mentions obsolete JDK APIs, like data structures
like Vector and Stack. But mostly LinkedList, which should usually be
replaced by ArrayList. And this is what this commit largely does.
JDK-4774077: https://bugs.openjdk.org/browse/JDK-4774077
1: https://issuetracker.google.com/issues/369219141
ExtensionElement is now a marker interface that requires all
implementation non-abstract classes to carry a static final QNAME
field (of type QName). This is verified by a new unit test.
Also FullyQualifiedElement is renamed to simply XmlElement. XmlElement
is used over ExtensionElement when implementing classes do not
statically know the qualified name of the XML elements they
represent. In general, XmlElement should be used sparingly, and every
XML element should be modeled by its own Java class (implementing
ExtensionElement).
This also lifts a bunch of logic from smack-websocket-okhttp into
smack-websocket. Furthermore, the following subprojects require now
Java 11:
- smack-integration-test
- smack-omemo-signal-integration-test
- smack-repl
- smack-websocket-java11
Related tracking issue: SMACK-835
This also means that smack-integration-test needs to declare a
dependency on Guava, which was previously available as transitive
dependency of 'reflections' 0.9.11.
This also resulted in a refactoring of the Providers and parsing
Exceptions. NumberFormatException and ParseException can now be thrown
directly, the wrapping in a SmackParsingException is down at a higher
layer, i.e. in AbstractProvider.
In order to truely stay Java 8 compatible, declaring a source and
target compatiblity is not sufficient. Source compatiblity means that
the input, i.e. the code written in Java is compatible with that
particular version of the Java Language Specification (JLS). And
target compatibitliy means that the produced Java bytecode is
compatible with that particular version of the Java Virtual Machine
Specificiation (JVMS).
But there is actually a third dimension: the runtime
library (rt.jar). If signatures of methods change over java releases
within the runtime library, then the produced bytecode, may contain
calls to methods that do not exist with that exact same signature in
older java versions.
For example the family of Buffer subclasses changed the return value
of certain functions, for example flip() to not return Buffer, but the
concrete type of the current instance, e.g. CharBuffer.
If we compile now with a newer JDK, where the return type is
CharBuffer and not Buffer, then executing on an older JDK, where the
return type is Buffer, then we get java.lang.NoSuchMethodError(s)
thrown at us.
Fixes SMACK-651.
